Uncertainty-Aware Map-Space Dynamics Models for Manipulation-Enhanced Mapping
A critical task in many robotic applications is acquiring and consistently updating an accurate and detailed model of the environment to plan and execute diverse actions in it. This is especially true in interactive scenes, where objects can be moved by the robot or humans. To efficiently maintain a map representation of such environments, one solution is to apply Next Best Viewpoint planning (NBV) [11] to reduce the uncertainty about the environment while minimizing the required number of observations to update the map. However, in confined and cluttered scenes, e.g. shelves, observing all objects in the scene is not always possible due to occlusions, leading to an incomplete representation and, consequently, difficulties in searching and retrieving desired objects…
- Published in:
IEEE/RSJ IROS 2024 Workshop - Type:
Inproceedings - Authors:
Dengler, Nils; Marques, Joao Marcos Correia; Zaenker, Tobias; Kalagaturu, Vamsi; Wang, Shenlong; Bennewitz, Maren; Hauser, Kris - Year:
2024 - Source:
https://lfm2024.github.io/papers/dengler_manipulation_enhanced.pdf
Citation information
Dengler, Nils; Marques, Joao Marcos Correia; Zaenker, Tobias; Kalagaturu, Vamsi; Wang, Shenlong; Bennewitz, Maren; Hauser, Kris: Uncertainty-Aware Map-Space Dynamics Models for Manipulation-Enhanced Mapping, IEEE/RSJ IROS 2024 Workshop, 2024, https://lfm2024.github.io/papers/dengler_manipulation_enhanced.pdf, Dengler.etal.2024b,
@Inproceedings{Dengler.etal.2024b,
author={Dengler, Nils; Marques, Joao Marcos Correia; Zaenker, Tobias; Kalagaturu, Vamsi; Wang, Shenlong; Bennewitz, Maren; Hauser, Kris},
title={Uncertainty-Aware Map-Space Dynamics Models for Manipulation-Enhanced Mapping},
booktitle={IEEE/RSJ IROS 2024 Workshop},
url={https://lfm2024.github.io/papers/dengler_manipulation_enhanced.pdf},
year={2024},
abstract={A critical task in many robotic applications is acquiring and consistently updating an accurate and detailed model of the environment to plan and execute diverse actions in it. This is especially true in interactive scenes, where objects can be moved by the robot or humans. To efficiently maintain a map representation of such environments, one solution is to apply Next Best Viewpoint planning...}}
